Protein-protein interactions are essential to almost all biological processes. Engineered proteins with novel binding properties are used as therapeutics and are important tools for cellular and biomolecular research. The objective of the research described here is to develop and test computational methods for manipulating the affinities and specificities of protein-protein interactions. This is an ongoing project and during the previous funding period we made encouraging progress in several areas including the high accuracy design of new protein complexes. Here we propose three specific aims that expand on previous successes while also tackling unsolved issues in interface design: (1) increasing binding affinities by the addition of new interface contacts, (2) specificity design through more rigorous negative design and (3) the design of buried hydrogen bond networks at protein interfaces. Each of our aims takes advantage of protocols we have added to the Rosetta molecular modeling program for optimizing the sequence, docked position and backbone conformations of interacting partners. All three aims include innovative aspects. In Aim 1 we will use a new strategy for the de novo design of proteins to stabilize binding epitopes by embedding them in folded proteins. In Aim 2 we will simultaneously model alternative binding conformations when designing sequences that favor one binding partner over another. In Aim 3 we will combine a new hydrogen bonding potential we have developed with sampling protocols that more tightly couple side chain and backbone sampling to create hydrogen bond networks at protein-protein interfaces. As test cases for our new computational protocols, we will design binders that competitively inhibit signaling molecules and enable biosensors that are less likely to perturb the system they are monitoring. Binding affinity measurements, site-directed mutagenesis and high-resolution structure determination will be used to evaluate the computational predictions. By pursuing this project we will expand the capabilities of computational protein design and test our understanding of the primary determinants of affinity and specificity at protein-protein interfaces.